shrg7/MolmoBot-Pi0-DROID-equad-fullft

MolmoBot-Pi0-DROID-equad (full fine-tune)

Full fine-tune of MolmoBot-Pi0 on the equad DROID dataset (shrg7/droid-equad-stage-1). This is the step-4000 checkpoint (training was still in progress at upload time).

Contents (inference-ready; optimizer state excluded):

• model.safetensors — full model weights
• metadata.pt, train_config.pkl — training/config metadata
• assets/droid_equad/norm_stats.json — action/proprio normalization stats

Sibling LoRA checkpoint: shrg7/MolmoBot-Pi0-DROID-equad.